Class SecureAggregationOptions

Namespace

AiDotNet.Models.Options

Assembly

AiDotNet.dll

Configuration options for secure aggregation in federated learning.

public class SecureAggregationOptions

Inheritance

object

SecureAggregationOptions

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

For Beginners: Secure aggregation hides each client's update from the server so the server can only see the final combined result. These options let you pick how secure aggregation should behave, including whether the protocol can handle clients dropping out mid-round.

Properties

Enabled

Gets or sets whether secure aggregation is enabled.

public bool Enabled { get; set; }

Property Value

bool

Remarks

This is an alternative to UseSecureAggregation. If either is enabled, secure aggregation is used.

MaxDropoutFraction

Gets or sets the maximum fraction of selected clients that may drop out while still completing the round.

public double MaxDropoutFraction { get; set; }

Property Value

double

Remarks

Used only when MinimumUploaderCount is not explicitly set. A value of 0.2 means "tolerate up to 20% dropouts" for the default minimum uploader calculation.

MinimumUploaderCount

Gets or sets the minimum number of clients that must upload masked updates for the round to succeed.

public int MinimumUploaderCount { get; set; }

Property Value

int

Remarks

If this is 0 or less, an industry-standard default is computed based on the number of selected clients.

Mode

Gets or sets which secure aggregation mode is used.

public SecureAggregationMode Mode { get; set; }

Property Value

SecureAggregationMode

ReconstructionThreshold

Gets or sets the reconstruction threshold used by dropout-resilient secure aggregation.

public int ReconstructionThreshold { get; set; }

Property Value

int

Remarks

If this is 0 or less, a default is chosen based on MinimumUploaderCount. The threshold is the minimum number of clients that must participate in the unmasking step to reconstruct missing masks when other clients drop out after uploading.